Simpson’s paradox, or the Yule-Simpson effect, is a statistical finding that the observed effect in several subgroups may be reversed when the groups are combined for analysis (

. In a UK case-control study (Ibid: 1295–7, see also editorial, ibid: 1283–4) the administration of parenteral penicillin by general practitioners to children with meningococcal disease was associated with a 7.4-fold increase in risk of death. The most likely explanation seems to be that the children given penicillin had more severe disease but it is possible that penicillin given before admission could be harmful. Simpson’s paradox might explain the observation that when all children with meningococcal disease were included in the analysis penicillin was associated with a slight reduction in risk of death but when the analysis was restricted to children whose general practitioner had suspected the diagnosis there was a marked increase in risk of death. The latter analysis excluded children who could not, or would not, have been given penicillin before admission. The authors of the case-control paper call for a randomised controlled trial but such a trial might be difficult to organise.

A study in Glasgow (

) has indicated that many people believe that bile in a newborn infant’s vomit is usually yellow. Forty-seven general practitioners, 29 special care baby unit nurses, 48 postnatal midwives and 41 mothers were asked to identify the colour of a bile-containing vomit from a …